Grid computing has been developed extensive in recently years and is becoming an important platform for high performance computing in scientific areas. Grid performance evaluation is an important approach to improve the performance of grid systems and applications. However, few work has been conducted in grid performance evaluation due to a lot of reasons like lack, of appropriate grid performance metrics, complexity of the grids, etc. In this paper, we analyze the performance metrics like response time and system utilization in the computational grid environment. We argue that instead of calculating the system utilization in the traditional, a concept of relative grid utilization, which describes how close is a single grid application performance to the performance of the same application submitted without grid middleware. We also discuss the utilization for the grid systems processing a number of jobs. In our experiments, we use NPB/NGB to evaluate the APSTC cluster grid and NTU campus grid performance, especially the overhead of SGE and Globus. Our results show that the overhead of grid middleware turns to be ignorable when the job size grows, and the characteristics of the grid applications affect a lot on utilization of computational grids.